Effects of moisture on automatic textile fiber identification by NIR spectroscopy

Lahti UAS has recently acquired a textile identifying and sorting unit REISKAtex® in order to develop identification analytics for different textile fibers. This article evaluates the effect of various humidity conditions in near infrared (NIR) spectrum of three different textile fiber materials, namely cotton, wool, and polyester.

Authors: Jussi Salin and Lea Heikinheimo

Introduction

Textile recycling has a significant environmental impact. In Finland, 71.2 million kg of textiles is removed from use each year (Dahlbo et al. 2015, 41). Various existing and new recycling processes for textile fibers depend on the purity and the right type of fiber material for each recycling process, because wrong materials create interference (Schmidt et al. 2016, 9; Fontell & Heikkilä 2017, 36). Automatic sorting could allow a larger portion of the textile waste flow to be processed into new fibers, if the fiber material contents of the recyclable textiles can be identified in order to send each textile for appropriate processing. In automatic sorting, a NIR analyzer could be used to identify the fiber materials of the recyclable textiles.

Water is known to be a significant variable in NIR spectroscopy, and therefore it could affect the automatic identification result of a NIR analyzer (Smith 2011, 16). Water absorption is used to determine the amount of water absorbed in textile materials under specified conditions. Factors affecting water absorption of a fabric are type of textile fiber, fabric structure, temperature, and length of exposure.

The analyzer used in this study is attached to a sorting unit located at Lahti UAS. This study is part of the Telaketju project. Telaketju is a co-operation network in Finland, which promotes circular economy by creating improvements both in recycling processes and in the flow of materials between companies. Telaketju is coordinated by VTT and Lounais-Suomen Jätehuolto Oy. The storage conditions of discarded textiles have raised concerns, including the effects of absorbed moisture. Developing automatic textile sorting is one key area of improvement of recycling. (Fontell & Heikkilä 2017, 31; Telaketju 2018.)

Testing methods and equipment

All fabrics used in the test have been stored in a normal room at the faculty, which has been at about 19 % relative humidity (RH) and 19 °C temperature throughout the experiment. The fabrics that are used for moisture testing are dried in a UT 12 drying cabinet by Kendro Laboratory Products at 104 °C. They are being dried till their weight stabilizes. An A&D GF-3000 digital scale is used for weighing the samples. Dry weights of the test fabric pieces can be obtained at this point. Next, the test fabrics are placed in various conditions, where they absorb air moisture till their weight no longer increases. The various moisture conditions are generated either by an ARC-500 weather cabinet by ArcTest company, or in a special room that has a Conairr CP3 air moisturizer and a temperature-controlled Glamox 200 radiator. (Salin 2018, 64-65.)

Between each tested moisture condition, the test fabrics are dried again to eliminate the hysteresis effect that occurs in textile fibers. If the fabric was not dried, it would gain slightly more moisture in a moist condition for being already in a more “open” state. In standard test methods, conditioning should always begin in the dry state (Collier & Epps 1999, 64).

NIR spectrums are obtained with NIRS Analyzer Pro by Metrohm AG, which is accompanied by Vision software. The software is used for gathering spectrums of textile samples, plotting them as graphs, and for creating an identification library. The identification library is trained with numerous samples of all textile fiber material groups chosen for the test. After verifying the library, it is then possible to attempt automatic identification of the test samples in their different moisture states, to report if identification fails at certain known amounts of moisture. The spectral range of the analyzer is between 1100 nm and 1650 nm (Metrohm AG 2017).

Fabric samples

Textile samples are taken from the textile library of Lahti UAS, which has collected fabrics of various fiber materials by various textile and fiber manufacturers. A total of 65 cotton fabrics, 9 wool fabrics and 178 polyester fabrics were chosen for training the identification library in Vision software (Salin 2018, 31).

One separate fabric piece of each fiber material is chosen for moisture testing. The structure of all three fabrics is plain weave (Salin 2018, 66).

Effects on fabric weight

The digital scale reports weights with 0.01 g accuracy when test fabrics are measured multiple times in a row. After weighing the test fabrics in each condition and calculating how much their weight has changed from dry weight, a graph is drawn (see Figure 1). The weight of wool is greatly increased by air humidity, it therefore being the most hydrophilic fiber material in the test, whereas cotton shows only relatively small increases. Polyester appears to be unaffected by humidity.

Figure 1. Measured water content increase of each test sample in different conditions next to commercial moisture regain coefficients located at 65.0 % RH and 20.0 °C (Salin 2018, 69).

By knowing dry weights of the test fabrics, it is possible to calculate water content regain coefficients of each measured condition. The measured coefficients can be compared to commercial moisture regain coefficients listed in the SFS 4876 standard. Coefficients of the standard are specified for 65.0 % ± 4.0 % RH and 20.0 °C ± 2.0 °C standard atmosphere condition of the SFS-EN ISO 139/A1 standard (SFS-EN ISO 139/A1). In Figure 1, the commercial moisture regain coefficients are drawn at 65.0 % RH as dots, next to the measured coefficients connected by lines. The commercial moisture regain coefficients are reasonably in line, except for polyester. The polyester test piece does not gain weight to an extent that can be measured by the digital scale even at 85 % RH, but commercial moisture regain expects it to gain 1.50 % more weight at 65.0 % RH (SFS 4876). That would be an 0.2 g increase to the 13.2 g dry weight of the test piece.

Effects on spectrum

Spectrums are gathered of each condition and test fabric, shown in Figure 2. Judging from the weight, wool and cotton absorb water content from air humidity, while polyester appears unaffected. The same effect can be seen in how the spectrum of polyester appears relatively unchanged, while wool and cotton have definite changes by absorbed water. The first overtone of water (H2O) causes a peak at 1460 nm, and the first overtone of hydroxide (OH), which is bundled in small amounts along water moisture, causes a peak at 1600 nm (Davies 2017). The more moisture the fabrics have absorbed, the greater the change in the spectrum. Cotton has relatively small changes because it is less hydrophilic than wool. Because of this, as an additional demonstration, the cotton test fabric is held in running water and then a spectrum is acquired again, which can also be seen in Figure 2.

Figure 2. Non-pretreated NIR absorbance spectrums of cotton, wool, and polyester test fabrics, at 1100-1650 nm, as water content changes in different humidity conditions (Salin 2018, 70-71).

To produce one spectrum, NIR sampling is done 32 times by the analyzer, in order to reduce noise. Spectrums in Figure 2 are averaged.

Effects on automatic identification

When running automatic identification for the test fabrics in Vision software, all spectrums are correctly identified without an error, except the experimental cotton sample that is directly soaked in running water. No other spectrums are ambiguous, non-identified nor mistaken as wrong material (Salin 2018, 72.)

The identification algorithm in use is Correlation in Wavelength Space, with threshold value of 0.73. The threshold value is forked by trial-and-error and determined by result of zero failures as the most optimal for this identification library. Calculation of 2nd derivate and Standard Normal Variate (SNV) are used as spectral pre-treatments, as they perform adequately in verification. (Salin 2018, 41-43.)

Conclusion

Textile recycling can have a large environmental effect. It has been estimated, that for example in Scandinavia textiles create the largest environmental impact after food, housing, and mobility (Schmidt et al. 2016, 7). By automatic sorting, recycling can be improved as more textiles can be sent for appropriate processing by their known chemical composition. This enables the use of both mechanical and chemical fiber recycling processes that are unique to each fiber material of sorted textiles. Water content in textiles could however pose a problem for automatic identification with NIR analysis, which is used to make the sorting decisions (Smith 2011, 16). The experimental results of this study answer to some questions about the practical moisture sensitivity in automatic textile identification by NIR analysis. Furthermore, to make the results practical, the same NIR analyzer unit was used in this study that is being used in the REISKAtex® sorting unit of LUAS, which is a model that can be used on industrial scale.

When the identification library was trained with samples stored at 19 °C and 19 % RH conditions, it was still possible to correctly identify textiles that were dry, as well as textiles that had been kept at 85 % RH of 20 °C (Salin 2018, 72). This wide range of acceptable changes in water content was the major finding of this study. Wool fabric was the most hydrophilic fabric, measured by water absorption, and it also had the greatest changes in spectrum, therefore being the most moisture sensitive textile material for NIR identification. Cotton fabric was also hydrophilic, but it was a less sensitive material because of smaller changes in both spectrum and weight. Polyester fabric did not gain water absorption in measurable amounts and had no noticeable changes in spectrum, being hydrophobic and the least moisture sensitive material for NIR identification.

Considering the experiments discussed in this article, it would appear that humidity does not pose an obstacle for automatic identification of single fiber cotton, wool, and polyester textiles. Every test fabric piece was identified correctly in all intended conditions of the experiment. It should be noticed, though, that the experiments did not go beyond 85 % relative humidity of 20 °C.

References

Collier, B. & Epps, H. 1999. Textile Testing and Analysis. New Jersey: Prentice-Hall, Inc.

Dahlbo, H., Aalto, K., Salmenperä, H., Eskelinen, H., Pennanen, J., Sippola, K. & Huopalainen, M. 2015. Tekstiilien uudelleenkäytön ja tekstiilijätteen kierrätyksen tehostaminen Suomessa. [Online document]. Helsinki: Ympäristöministeriö. [Cited 16 May 2018]. Available at: https://helda.helsinki.fi/bitstream/handle/10138/155612/SY_4_2015.pdf

Davies, A. 2017. An introduction to near infrared (NIR) spectroscopy. [Cited 16 May 2018]. Available at: http://www.impublications.com/content/introduction-near-infrared-nir-spectroscopy

Fontell, P. & Heikkilä, P. 2017. Model for circular business ecosystem for textiles. [Online document]. Espoo: VTT.  VTT Technology 313. [Cited 16 May 2018]. Available at: http://www.vtt.fi/inf/pdf/technology/2017/T313.pdf

Metrohm AG. 2017. NIRS Analyzer PRO – DirectLight/NonContact. [Cited 16 May 2018]. Available at: https://www.metrohm.com/en-gb/products-overview/process%20analyzers/applikon%20nirs%20pro/A629281130

Salin, J. 2018. Automatic Identification of Textiles with NIR-spectroscopy. Master’s thesis. Lahti University of Applied Sciences, Faculty of Technology. Lahti.

Schmidt, A., Watson, D., Askham, C. & Brunn Poulsen, P. 2016. Gaining benefits from discarded textiles. LCA of different treatment pathways. [Online document]. Denmark: Nordic Council of Ministers. TemaNord 2016:537. [Cited 16 May 2018]. Available at: https://norden.diva-portal.org/smash/get/diva2:957517/FULLTEXT02.pdf

SFS 4876. 1987. Tekstiilit. Kuitusisällön ilmoittaminen. Helsinki: Finnish Standards Association SFS.

SFS-EN ISO 139/A. 2005. Textiles. Standard atmospheres for conditioning and testing. Helsinki: Finnish Standards Association SFS.

Smith, B. 2011. Fundamentals of Fourier Transform Infrared Spectroscopy. Boca Raton: CRC Press.

Telaketju. 2018. Telaketju ­– Mikä se on? [Cited 16 May 2018]. Available at: https://telaketju.turkuamk.fi/mita_telaketju_tekee/

Authors

Jussi Salin is a Master’s Degree student at Lahti UAS in the  Programme in Smart Industries and New Business Concepts.

Lea Heikinheimo, D.Sc. (Tech), is a principal lecturer at Lahti UAS, Faculty of Technology, in the Degree Programme in Process and Materials Technology and in the Master’s Degree Programme in Smart Industries and New Business Concepts.

Published 24.5.2018

Illustration: Oona Rouhiainen

Reference to this publication

Salin, S. & Heikinheimo, L. 2018. Effects of moisture on automatic textile fiber identification by NIR spectroscopy. LAMK RDI Journal. [Electronic journal]. [Cited and date of citation]. Available at: http://www.lamkpub.fi/2018/05/24/effects-of-moisture-on-automatic-textile-fiber-identification-by-nir-spectroscopy/

Yksi kommentti artikkeliin ”Effects of moisture on automatic textile fiber identification by NIR spectroscopy

Vastaa

Sähköpostiosoitettasi ei julkaista. Pakolliset kentät on merkitty *